IMAGE-BASED PRODUCTS RECOMMENDATION SYSTEM

Abstract

The research team has developed a system based on a CNN network to suggest products similar to the one being searched. The system operates with a test sample consisting of 40 input images and 30 different objects, yielding excellent results. Expanding the model, the system can handle an unlimited amount of data, providing fast image-based search results (running on a CPU-based computer, with search results delivered within 300ms).